Image Sharpening Detection Based on Difference Sets

Abstract

Image sharpening is one of the basic operations used to improve the visual effect of images. Image editing makes it impossible to confirm the authenticity of an image, and the purpose of image forensics is to detect whether an image has been artificially edited. To solve this problem, a forensic algorithm based on global image pixel values is proposed. Twelve difference sets composed of first-order and second-order differences in different directions are used as the image feature to reveal the difference between image pixels. Additionally, the feature is used to determine whether the image has been sharpened or not. To verify the performance of the algorithm, a series of experiments are performed on different formats of image datasets. Compared with the existing forensic algorithms based on traditional machine learning, the detection accuracy of the proposed method is greatly improved.